Day The Four

library(tidyverse)

steam <- read_csv("raw_data/steam_checkpoint_2.csv")

Rows: 26564 Columns: 27── Column specification ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Delimiter: ","
chr  (10): name, developer, publisher, controller_support, multiplayer, categories, genres, steamspy_tags, general_rating, owners
dbl   (9): appid, required_age, achievements, positive_ratings, negative_ratings, average_playtime, median_playtime, price, total_reviews
lgl   (7): free_to_play, virtual_reality_support, steam_workshop, singleplayer, windows_support, mac_support, linux_support
date  (1): release_date
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.Error in exists(cacheKey, where = .rs.WorkingDataEnv, inherits = FALSE) : 
  invalid first argument
Error in assign(cacheKey, frame, .rs.CachedDataEnv) : 
  attempt to use zero-length variable name

steam

steam %>% 
  select(name, owners)

#For the purposes of making a model, what would we want the target to be? If we were to create a column called “successful”, what variables would it look for?

steam %>%
  separate(col = steamspy_tags, sep = ";", into = c("tag_1", "tag_2", "tag_3")) %>% 
  pivot_longer(cols = c("tag_1", "tag_2", "tag_3"), values_to = "tags", names_to = "drop_me") %>% 
  drop_na(tags) %>% 
  group_by(tags) %>% 
   count(tags) %>% 
   arrange(desc(n)) 

Warning: Expected 3 pieces. Missing pieces filled with `NA` in 2551 rows [1283, 3566, 3669, 3711, 4035, 4146, 4457, 4733, 4822, 5019, 5140, 5142, 5264, 5361, 5378, 5497, 5553, 5556, 5570, 5756, ...].

# creating new logical columns for use in the model - we cant have every genre/tag, so a selection of a few will have to do
# Preferably ones without a lot of overlap
# Not including Indie just now, because anything developed by Dinkey game would be Indie by default - It's a tag that doesnt really describe anything about the game
steam <- steam %>% 
  mutate(is_open_world = ifelse(str_detect(steamspy_tags, "Open World"), TRUE, FALSE),
         is_rogue_like = ifelse(str_detect(steamspy_tags, "Rogue-like"), TRUE, FALSE),
         is_metroidvania = ifelse(str_detect(steamspy_tags, "Metroidvania"), TRUE, FALSE),
         is_visual_novel = ifelse(str_detect(steamspy_tags, "Visual Novel"), TRUE, FALSE),
         has_great_soundtrack = ifelse(str_detect(steamspy_tags, "Great Soundtrack"), TRUE, FALSE),
         is_shooter =ifelse(str_detect(steamspy_tags, "Shooter") | str_detect(steamspy_tags, "FPS") | str_detect(steamspy_tags, "Third-Person Shooter"), TRUE, FALSE),
         is_rpg = ifelse(str_detect(steamspy_tags, "RPG"), TRUE, FALSE),
         has_female_protagonist = ifelse(str_detect(steamspy_tags, "Female Protagonist"), TRUE, FALSE),
         is_sports = ifelse(str_detect(steamspy_tags, "Sports"), TRUE, FALSE),
         is_simulation = ifelse(str_detect(steamspy_tags, "Simulation"), TRUE, FALSE),
                                .after = multiplayer) 

Error in exists(cacheKey, where = .rs.WorkingDataEnv, inherits = FALSE) : 
  invalid first argument
Error in assign(cacheKey, frame, .rs.CachedDataEnv) : 
  attempt to use zero-length variable name

steam <- steam %>% 
  mutate(is_positive = ifelse(str_detect(general_rating, "Positive"), TRUE, FALSE),
         .after = linux_support)

Error in exists(cacheKey, where = .rs.WorkingDataEnv, inherits = FALSE) : 
  invalid first argument
Error in assign(cacheKey, frame, .rs.CachedDataEnv) : 
  attempt to use zero-length variable name

For the purpose of my model over in Python land, i’m turning Multiplayer into just a logical to see if it changes anything

steam %>% 
  distinct(multiplayer)

steam <- steam %>% 
  mutate(multiplayer = ifelse(str_detect(multiplayer, "No multiplayer"), FALSE, TRUE)) 

Error in exists(cacheKey, where = .rs.WorkingDataEnv, inherits = FALSE) : 
  invalid first argument
Error in assign(cacheKey, frame, .rs.CachedDataEnv) : 
  attempt to use zero-length variable name

#ok fine im adding indie to see what happens 
steam <- steam %>% 
  mutate(is_indie = ifelse(str_detect(steamspy_tags, "Indie"), TRUE, FALSE)) 

Error in exists(cacheKey, where = .rs.WorkingDataEnv, inherits = FALSE) : 
  invalid first argument
Error in assign(cacheKey, frame, .rs.CachedDataEnv) : 
  attempt to use zero-length variable name

steam %>% 
  filter(multiplayer == FALSE)

# what if we bumped up the threshold? What if mostly positive was no longer acceptable?

steam_but_harsher <- steam %>% 
  mutate(is_positive = case_when(
    general_rating == "Extremely Positive" | general_rating == "Positive" ~ TRUE,
    TRUE ~ FALSE
  )) 

write_csv(steam, "clean_data/steam_for_model.csv")
write_csv(steam_but_harsher, "clean_data/harsher_steam_for_model.csv")

---

For now, im gonna do some text stuff using backloggd reviews

backloggd_reviews <- read_csv("clean_data/backloggd_reviews.csv")

Rows: 1509 Columns: 8── Column specification ─────────────────────────────────────────────────────────────
Delimiter: ","
chr (7): title, summary, reviews, genre_tag, genre_tag_2, genre_tag_3, genre_tag_4
dbl (1): rating
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

# separating reviews into their own columns so i can manage them
reviews_separated <- backloggd_reviews %>% 
  separate_longer_delim(cols = "reviews", delim = "',\ \'") %>% 
  separate_longer_delim(cols = "reviews", delim = "\',\ \"") %>% 
  separate_longer_delim(cols = "reviews", delim = "\",\ \'") %>%  
  separate_longer_delim(cols = "reviews", delim = "\",\ \"") %>% 
  unique() %>% 
  # removing all the opening and closing rubbish
  mutate(reviews = str_remove_all(string = reviews, pattern = "\\[\""),
         reviews = str_remove_all(string = reviews, pattern = "\\[\'"),
         reviews = str_remove_all(string = reviews, pattern = "\"\\]"),
         reviews = str_remove_all(string = reviews, pattern = "\'\\]")) 

reviews_separated %>% 
  group_by(title) %>% 
  count(title)

lets split this into ratings, like i did with the other backloggd thing

reviews_separated <- reviews_separated %>% 
  mutate(rating_range = case_when(
    rating < 5 & rating >= 4 ~ "4+",
    rating < 4 & rating >= 3 ~ "3 to 4",
    rating < 3 & rating >= 2 ~ "2 to 3",
    rating < 2 & rating >= 1 ~ "1 to 2",
    rating < 1 ~ ">1",
    TRUE ~ "No rating"
  ))

Before doing text analysis, lets convert everything to lower case - this might take away some context via capitalisation, but we only have 6 reviews per game so gotta work with what we have

reviews_lower_separated <- reviews_separated %>% 
  mutate(reviews = tolower(reviews)) #%>% 
  #filter(reviews %in% c("í", "ñ", "é", "á", "ó", "á") == FALSE)

library(textdata)
library(tidytext)

de_stopwords <- tibble(word = stopwords("de"))

Error in stopwords("de") : could not find function "stopwords"

reviews_3_to_lowest_stop <- reviews_lower_separated %>% 
  filter(rating_range == "2 to 3" | rating_range == "1 to 2" | rating_range == ">1") %>% 
  select(-summary) %>% 
  unnest_tokens(input = reviews, output = word) %>% 
  anti_join(filter(stop_words, lexicon == "SMART")) %>% 
  anti_join(game_stop_words) %>% 
  anti_join(pt_stopwords) %>% 
  anti_join(de_stopwords) %>% 
  count(word, sort = TRUE)

Joining with `by = join_by(word)`Joining with `by = join_by(word)`Joining with `by = join_by(word)`Joining with `by = join_by(word)`

bigrams might be more appropriate for bad reviews

bigrams_3_to_lowest_stop <- reviews_lower_separated %>% 
  filter(rating_range == "2 to 3" | rating_range == "1 to 2" | rating_range == ">1") %>% 
  unnest_tokens(bigram, reviews, token = "ngrams", n = 2) %>% 
  count(bigram, sort = TRUE) %>%  # count bigrams
  separate(bigram, into = c("word1", "word2"), sep = " ") %>% # split bigrams into two seperate columns
  anti_join(stop_words, join_by("word1" == "word")) %>%  # check if word1 of bigram is a stop word
  anti_join(stop_words, join_by("word2" == "word")) %>% 
  anti_join(pt_stopwords, join_by("word1" == "word")) %>% 
  anti_join(pt_stopwords, join_by("word2" == "word")) %>% 
  anti_join(game_stop_words, join_by("word1" == "word")) %>% 
  anti_join(game_stop_words, join_by("word2" == "word")) %>% 
  anti_join(de_stopwords, join_by("word1" == "word")) %>% 
  anti_join(de_stopwords, join_by("word2" == "word")) %>% 
  drop_na() %>% 
  unite(col = "bigram", word1:word2, sep = " ", remove = TRUE) %>% 
  anti_join(game_stop_bigrams)

Joining with `by = join_by(bigram)`

well fine lets remove game names as well

ok that still kinda sucks

  ggwordcloud(words = bigrams_3_to_5_no_stop$bigram, freq = bigrams_3_to_5_no_stop$n, random.color = TRUE, colors =  c("#ff595e", "#ffca3a", "#8ac926", "#1982c4", "#6a4c93"))

ok what if we try some of that tf-idf stuff

``